Menu

Long silence increases misrecognition

Help
2010-07-14
2012-09-22
  • Krisztian Loki

    Krisztian Loki - 2010-07-14

    Hi All,

    recently I tried to implement some basic barge-in capability but quickly
    discovered that both PocketSphinx and Sphinx3 have the same bug: if the
    utterance begins with a long silence (a couple of seconds), then the
    recognition is totally bogus. An all-silence utterance is correctly recognised
    as noise/silence. Have any of you experienced something similar?

     
  • Nickolay V. Shmyrev

    This is not a bug, you are not using endpointer, that's it.

     
  • Krisztian Loki

    Krisztian Loki - 2010-07-14

    Indeed I'm not, I don't even know what it is. So a simple pocketsphinx_batch
    on a raw file starting with silence won't give the expected result? What do I
    have to do to make it work?

     
  • Nickolay V. Shmyrev

    pocketsphinx_batch is supposed to decode short files without silences.

    If you need to decode long files with silences you need to use pocketsphinx
    API differently. You need to filter silences using cont_ad functions from
    sphinxbase. You can find example of API usage looking on
    pocketsphinx_continuous. There were also many threads about that in this
    forum.

     

Log in to post a comment.

Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.